From September 2002 through February 2008, 1,691 patients with NHL were enrolled onto the study; this does not include chronic lymphocyte leukemia/small lymphocyte leukemia, which is being reported elsewhere. Of the 1,691 patients, 701 were excluded because they did not have a serum sample available and/or drawn within 120 days of diagnosis and seven were excluded for missing clinical or outcome data, leaving a total of 983 subjects available for analysis.
Of the 983 patients with NHL, the mean age at diagnosis was 62 years (range, 19 to 94 years) and 55% were male. Overall, 56 patients (5.7%) had severe insufficiency (< 10 ng/mL), 380 (38.7%) had mild to moderate insufficiency (10 to 24 ng/mL), and 547 (55.6%) were in the optimal range (25 to 80 ng/mL) for 25(OH)D; 436 (44%) were classified as 25(OH)D insufficient (combination of severe and mild to moderate insufficiency). 25(OH)D insufficiency was not correlated with age or sex, but was positively correlated with having a serum drawn during or post-treatment; residence in the six state region of the upper Midwest; diagnosis from March through May; PS higher than 1; and DLBCL or TCL subtypes (). Subtype-specific prognostic indices were also positively correlated with 25(OH)D insufficiency (), and this was most strongly driven by performance status within each of the subtypes (Appendix Table A2, online only). Disease stage was not correlated with 25(OH)D insufficiency for DLBCL, FL, or post-FL, but higher disease stage was correlated with 25(OH)D insufficiency in TCL (Appendix Table A2).
Demographic and Clinical Correlates of Vitamin D Levels and Prevalence of 25-Hydroxyvitamin D Insufficiency
During a median follow-up of 34.8 months (range, 0.5 to 77 months), there were 404 events and 193 deaths, of which 168 were attributed to lymphoma. For DLBCL, 52% of the patients were 25(OH)D insufficient, and as shown in , insufficient patients had inferior EFS (log-rank P = .005) and OS (log-rank P < .001); details on univariate HRs for all outcomes are available in Appendix Table A3 (online only). After adjusting for the IPI and treatment (), 25(OH)D insufficient patients continued to have inferior EFS (HR, 1.41; 95% CI, 0.98 to 2.04), LSS (HR, 2.16; 95% CI, 1.33 to 3.51) and OS (HR, 1.99; 95% CI, 1.27 to 3.13). The association of 25(OH)D levels with EFS (A) and OS (B) was mainly observed over the range of 15 to 25 ng/mL, and was relatively flat above 30 ng/mL.
Fig 1. Kaplan-Meier curves for 25-hydroxyvitamin D deficiency and (A) diffuse large B-cell (DLBCL) event-free survival (EFS), (B) DLBCL overall survival (OS), (C) T-cell lymphoma (TCL) EFS and (D) TCL OS. HR, hazard ratio; IPI, International Prognostic Index. (more ...)
Multivariable-Adjusted Hazard Ratios for 25-Hydroxyvitamin Insufficiency and Event-Free, Lymphoma-Specific, and Overall Survival
Fig 2. Estimated hazard ratios (solid line) and 95% CI (dotted lines) from multivariate models for 25-hydroxyvitamin D level (ng/mL) and (A) diffuse large B-cell (DLBCL) event-free survival, (B) DLBCL overall survival, (C) T-cell lymphoma (TCL) event-free survival, (more ...)
For TCL (which included PTCL and CTCL, see for details), 57% of the patients were 25(OH)D insufficient, and insufficient patients had inferior EFS (log-rank P = .003) and OS (log-rank P = .01). After adjustment for IPI (), 25(OH)D insufficient patients continued to have inferior EFS (HR, 1.94; 95% CI, 1.04 to 3.61), LSS (HR, 2.26; 95% CI, 0.99 to 5.17) and OS (HR, 2.38; 95% CI, 1.04 to 5.41). Inferior EFS (C) and OS (D) was observed among TCL patients in both the insufficient (< 25 ng/mL) and the lower end of the optimal range (25 to 80 ng/mL) of 25(OH)D levels.
The prevalence of 25(OH)D insufficiency for the remaining subtypes in ranged from 27% to 39%, and there were no associations of 25(OH)D insufficiency with EFS. However, with the exception of LSS in FL, HRs for LSS and OS for the remaining subtypes were all above 1, although the confidence intervals were wide, reflecting the small number of deaths in these subtypes.
Further adjustment of results in for season of diagnosis or residence in the upper six Midwest states did not materially alter the associations (< 10% change in HRs) reported in (data not shown).
All serum samples tested were drawn within 120 days of diagnosis, and from a physiologic perspective we would expect little change of 25(OH)D levels from chemotherapy over this short interval. Nevertheless, 34% of samples were drawn during or after initial therapy, and mean 25(OH)D levels and the prevalence of 25(OH)D insufficiency were modestly associated with timing of blood draw overall (). This overall association varied by subtype, and was seen only for TCL, MCL, and all other NHL, but not for DLBCL, FL, and post-FL (Appendix Table A2). Adjustment for timing of serum draw did not materially alter the associations in (data not shown). Further, when restricted to pretreatment serum samples (Appendix Table A4, online only), our basic findings held, although the CIs around the point estimates became unstable due to small numbers for subset analyses in the lymphoma subtypes.
One potential interpretation of the association of low 25(OH)D levels with inferior DLBCL and TCL prognosis is that patients with a larger tumor burden might have increased conversion of 25(OH)D to 1,25(OH)2
D due to increased 1-α-hydroxylase activity from the tumor, leading to artificially low serum 25(OH)D levels.30
As such, tumor size or aggressiveness might be a confounding factor. Although adjustment for IPI should remove most of the potential confounding, residual confounding remains a concern. Calcium levels were similar between 25(OH)D sufficient and insufficient patients with DLBCL and TCL (Appendix Table A5, online only). In addition, further adjustment of the results in for albumin-corrected-calcium (as a surrogate for tumor activity, since high 1-α-hydroxylase activity would be expected to also increase serum calcium levels) did not materially change the DLBCL results (Appendix Table A5). The results for TCL attenuated for EFS (HR, 1.35; 95% CI, 0.63 to 2.88) and to a lesser extent for LSS (HR, 2.00; 95% CI, 0.74 to 5.37), while OS remained similar (HR, 2.17; 95% CI, 0.82 to 5.76).
To further address this issue, we measured 1,25(OH)2D levels in patients with DLBCL and TCL. For DLBCL, patients in the lowest three quartiles of 1,25(OH)2D had inferior EFS, LSS, and OS (and ); this association was maintained when stratified by 25(OH)D sufficient versus insufficient. For TCL, patients below the median (too few patients to use quartiles) also had inferior outcomes in a pattern similar to that seen for DLBCL, although all estimates were imprecise and not statistically significant, likely due to small subject numbers ( and ). Overall, these data do not support the hypothesis that the association of lower 25(OH)D levels with poor prognosis in DLBCL and TCL is confounded by tumor production of 1-α-hydroxylase. Further, these data also indicate that there appears to be a direct association of lower 1,25(OH)2D levels with inferior outcome.
Kaplan-Meier curves for 1,25-dihydroxyvitamin D levels and (A) diffuse large B-cell (DLBCL) event-free survival (EFS), (B) DLBCL overall survival (OS), (C) T-cell lymphoma (TCL) EFS, and (D) TCL OS. 1,25-dihydroxyvitamin D levels in pg/mL.
Multivariable-Adjusted HRs for 1,25-Dihydroxyvitamin D Levels (overall, and stratified by 25-hydroxyvitamin D insufficiency) and Event-Free, Lymphoma-Specific, and Overall Survival